Prelimbic cortex-associated stress sensitisation and region-specific mitochondrial adaptations across preclinical approaches relevant to PTSD and depression.

Global implementation of artificial intelligence (AI) has been perceived as a means of improving capacity, capability, and efficiency in health systems, however with these prospects come inherent and unique risks which require appropriate assessment and mitigation. This is particularly the case in mental health which warrants specific consideration. The regulation of AI as a medical device (AIaMD) forms a key element of the approach to limit risks whilst optimising the benefits of this technology. The Medicines and Healthcare products Regulatory Agency (MHRA) regulates medicines and medical devices in the UK, including digital mental health technology (DMHT) that qualifies as AIaMD. The MHRA is leading on a programme exploring the regulation and evaluation of DMHT, funded by Wellcome, in partnership with the National Institute for Health and Care Excellence (NICE), England’s health technology assessment agency. As part of this project an expert symposium to explore AI in DMHT was convened by the MHRA in order to deliberate the critical topics in this field and engage with a diverse range of stakeholders. A series of presentations covering innovative research, future uses in mental health, developer perspectives, responsible use and considerations of equity, was followed by a roundtable discussion and an interactive workshop constructing a clinical trial for a hypothetical product. The 59 symposium members came from a range of backgrounds: academia (18 participants), industry including developers and manufacturers (12 participants), the healthcare sector including clinicians (9 participants), government and policy makers (3 participants), health technology assessment and evaluation agencies (3 participants), regulatory bodies (3 participants), research funding bodies (3 participants), patient representatives and lived experience experts (2 participants), charities (2 participants), independent researchers (2 participants), and non-profit organisations (2 participants). In this review the challenges and opportunities for AI in DMHT are outlined, illuminated by insights from the symposium. Themes were developed following the symposium after reflection by the lead author, and agreed by all authors, then shared with the symposium participants who presented. This was built upon by areas of consensus and significant viewpoints within the symposium, supported by the relevant literature, core regulatory principles and guidance. Whilst some areas fall outside of the remit of regulation (such as the use of generalised AI platforms for health and the use in administration or health system processes), they are nevertheless important to consider for a more comprehensive understanding of the AI ecosystem and the broader impacts to health services and society. Topics include the distinct diagnostic paradigms in mental health, generative AI applications in DMHT, medical device qualification as a vital risk management tool, ethical and equity considerations, and harmonisation of international regulatory systems. These reflections are intended to be valuable to regulators, government, policymakers, clinicians, academics, and developers alike.